The Agent Skills Directory

[SAFE]: The skill's primary logic is contained in a local C# source file (scripts/Program.cs) that uses the official OpenXML SDK to programmatically build document structures. It does not perform network requests, access sensitive system directories, or execute obfuscated code.
[SAFE]: External dependencies are limited to well-known, high-reputation libraries (DocumentFormat.OpenXml and Markdig) pulled from standard NuGet registries.
[SAFE]: Command execution is restricted to standard developer tooling for compilation (dotnet), document conversion (soffice), and image processing (pdftoppm). The provided AppleScript for PDF export uses standard macOS UI automation techniques and is well-documented with transparency regarding its locale requirements.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted markdown data as input for document generation, creating a potential attack surface for indirect prompt injection.
Ingestion points: Markdown content is ingested from local files and parsed into an Abstract Syntax Tree (AST) in scripts/Program.cs.
Boundary markers: No explicit delimiters are used for the markdown content; however, the skill operates as a structured transformer rather than a prompt-driven agent for the generation step.
Capability inventory: Capabilities include reading markdown files, writing .docx and .pdf files, and executing local conversion utilities.
Sanitization: Content is sanitized by the Markdig parser, which handles markdown as a syntax tree, preventing the execution of embedded scripts or commands within the document generation pipeline.

docx-creator